Reading strategy in Hong Kong school-aged children: The Development of Word Level and
Character Level Processing
Chu Mun-Ki Maggie & Leung Man-Tak
Division of Speech & Hearing Sciences, The University of Hong Kong, Hong Kong
ABSTRACT
This study uses frequency effect to investigate the development of mental representation of
Chinese disyllabic words in primary school-aged readers in Hong Kong. Unlike alphabetical
languages, written Chinese is a logographic system with character as basic unit. As most of the
Chinese words are composed of more than one character which can be viewed as single morpheme,
theoretically Chinese compound word can be read as a meaningful whole, its component
characters. Previous studies on Chinese reading (e.g. Zhang & Peng, 1992; Taft, Liu & Zhu, 1999;
Peng, Liu & Wang, 1999) showed that adults used both whole word level and character level
processing for reading Chinese multi-character words and there were interaction effect between
word and component character frequencies. In the present study, subjects were asked to read aloud
a list of two-character words which were controlled for word and component character frequencies
across grades. The percentage of correct responses was analysed by a three-way ANOVA.
Significant main effects of word frequency, component character frequency and grade were found.
Significant three-way interaction effects were also found, indicating that children as early as grade
one are able to make use of the word level and the component character level reading processing.
Lower graders tended to use the component character level reading processing more when reading
Chinese compound words, while higher graders tended to read words as whole units more.
KEY WORDS: multi-character words, word level reading, character level reading, reading
development.
INTRODUCTION
In alphabetic languages, the basic unit of orthography is letter (alphabet). Each letter corresponds
to a phoneme and reading is facilitated by the reader’s phonemic awareness. Chinese language is a
logographic writing system in which characters are the basic units. Each character (morpheme) has
its phonological representation which is a syllable and its meaning. Most Chinese words are multicharacter compound words. Zhou, Shu, Bi & Shi (1999) reported that two-character compound
words constitute about 73.6% of the words in Chinese. Compound words are composed of two or
more characters. Each of the component characters can usually be viewed as single morpheme,
which can combine with other characters to form other words. For example, the two-character
word 毛衣 (sweater) is composed of the first character 毛 (on the left) which means ‘fur’ and the
second character 衣 (on the right) which means ‘clothes’. Both component characters can combine
with other characters to form another words like 毛巾 (towel) and 外衣 (overcoat).
Unlike words in alphabetical language like English, there is no space demarcation for
Chinese compound words in text, which appears as a continuous concatenation of characters.
Theoretically, there are two possible ways of reading Chinese compound words: either reading it
character by character or reading it as an unanalyzed whole. For example, the compound word 毛
衣 /mou21 ji55/ can be read as a whole meaning ‘sweater’, or read, from left to right, as characters
毛 /mou21/ (‘fur’) and 衣 /ji55/ (‘clothes’). Similarly, two possible patterns of acquisition of
Chinese reading may exist. The first pattern suggests that when children are exposed to Chinese
characters as units of meaningful wholes, which were usually multi-character words, such as 電話
/tin22 wa35/ (telephone), rather than individual character 電 /tin22/ (‘electricity’) and 話 /wa35/
(‘word’), and consequently, the word level representation develops first. The second pattern is
based on the view that as there is no space demarcation for words in texts, children tend to read
character by character and they develop the character level representation prior to that of the word
level representation. The aim of the present study is to investigate the order of acquisition of the
word level and the component character level reading strategies of Chinese multi-character words
in children.
The presence of representations at component character level was hypothesized to facilitate
reading of new compound words (Liu and Peng 1997). With representation at single character
level, children can read aloud new compound characters through accessing the representations of
the component characters. Once they read aloud the component characters, they might be able to
understand the new word through accessing their corresponding phonological lexicons which has
already developed before six. On the other hand, the word level access was hypothesized to
increase the efficiency of reading Chinese multi-character words. When children came across
words with low frequency component characters, their ability to read aloud the words can be
enhanced by the access at word level for high word frequency words. As Chinese word processing
at both whole word level and component character level facilitated reading, it is assumed that both
levels of lexical representations are developed in advanced readers.
One of the well-known models of reading processing was the Augmented Addressed
Morphology (Caramazza, Laudanna & Romani, 1988) model (AAM model). The model
suggested that the whole word level and the character level of lexical representation were present
as independent processing levels. It assumed that the access to lexical representation was carried
out at whole word level for familiar words, and at morpheme level for unfamiliar words.
Zhang & Peng (1992) carried out two experiments to testify the AAM model (Caramazza
et al., 1988) and investigated processing levels of reading Chinese words in normal adults by
measuring the response time of subjects in a lexical decision task when disyllabic character stimuli
were presented. In the first experiment, they varied the word frequency and controlled the
frequency of component characters. Their prediction was that if there were word level
representation, word frequency effect would be obtained with the speed of word recognition on
high frequency words faster than that of low frequency words. Predicted results were obtained,
indicating that a whole word level processing is involved in adult’s reading processing of Chinese
words.
The second experiment by Zhang & Peng (1992) investigated the component character
frequency effect in adults’ Chinese reading. They varied the component character frequency while
the whole word frequency was controlled in a word decision task. The average response time for
HH (high component character frequency on left and right positions, high-high) words was found
to be shorter than HL (high-low) and LH (low-high) words. Component character frequency effect
was found and the results suggested that component character level processing was involved in
reading multi-character words. Taft, Huang & Zhu (1994), as reported by Peng, Liu & Wang
(1999), found similar results for component character frequency effect in a lexical decision task.
The response time for recognizing words with HH component character frequency was shorter
than that for words with HL and LH character frequency, which in turns was shorter than words
with LL component character frequency.
Based on the findings of the previous studies (Zhang & Peng, 1992; Taft & Zhu, 1995; Taft,
Liu & Zhu, 1999), Taft & Zhu (1995) and Taft, Liu & Zhu (1999) suggested a new model called
Multilevel Interactive-activation Framework to represent the reading processing of Chinese multicharacter words (see Figure 1). Unlike the independent processing at different representation levels
proposed by the AAM model, they proposed several interactive levels of orthographic
representations including the stroke level, radical level, component character (morpheme) level
and compound word level. It was hypothesised that activation passed up and back down the levels
of the interactive-activation system. Each level of orthographic representation, except the stroke
level, was linked to its semantic representation (meaning) and phonological representation. The
connection strength of the links to each level of representation depended on the frequency the link
was used. In other words, Chinese characters that were frequently shown at a particular level of
representation would have stronger link to that representation at the corresponding level, and thus
would be activated more readily.
Insert Figure 1 about here
Using the interactive-activation framework, Taft (1994) explained the difference between
reading low frequency compound word and high frequency compounds. When reading high
frequency words, representations at the word level was readily retrieved, therefore, little activation
was required at the component character level. For low frequency compound word, activation at
the word level was reduced, thus relatively more activation was required at the component
character level.
Peng, Liu & Wang (1999) investigated the interaction between whole word frequency and
component character frequency by measuring adults’ response latency in a word decision task.
They found significant main effects in both word frequency and component character frequency,
and interaction effect between word and component character frequencies. Contrary to Taft’s
(1994) predication, the component character frequency effect was found in high word frequency
condition only, instead of in low word frequency condition.
Peng et al. (1999) tried to explain the results by proposing that the connection strength
from component character level to the word level representation was stronger for high frequency
words. He went ahead explaining that the higher frequency the component characters exposed in a
word, the greater the association of the component characters with that word. As the connection
strength between the two levels was high in high frequency words, the access of representations at
word level was suggested to be affected by the component character frequency more.
In order to testify the two different put forward by Peng et al. (1999) and Taft (1994), the
whole word frequency and the component character frequency of Chinese two-character words
was varied in this experiment and the change of interaction between the frequencies across grades
was explored.
One possible explanation for the discrepancy between Peng et al.’s (1999) results and
Taft’s (1994) predication was related to the paradigm used by Peng’s et al.’s (1999) study. The
word decision task in Peng et al.’s (1999) experiment “forced” the subjects to access the word
level representations in order to make a lexical decision (decide whether the two-character stimuli
were words). The nature of the task induced the “so called” higher connection strength from
component character level to the word level representation for high frequency words, instead of
allowing subjects’ flexibility to process words at different levels of representations for different
word and component character frequencies. Given the above consideration, a reading aloud task
with percentage correct measurement were used in this study to prevent limiting the levels of
reading processing used by the subjects.
It was expected that there would be significant word frequency effect if the subjects
processed words at the word level. Activation thresholds of high frequency words were lower than
that of low frequency words and thus better reading performance in high frequency words was
predicted. If children processed words at the component character level representation, significant
component character frequency effect would be expected. The reading performance on HH words
would be better than HL/LH words which in turn would be better than LL words.
The interaction between the word frequency and the component character frequency across
grade would tell us more about the development of reading strategy across age. Significant
interaction effects between word frequency and component character frequency would be obtained
if children accessed representations of the words at both whole word level and component
character level when reading multi-character words. According to the AAM model and the
interactive-activation system, component character frequency effect would be expected for low
frequency but not for high frequency words. As the activation thresholds of low frequency words
were high, access through component character level was required to aid in reading low frequency
words. On the other hand, since it is easier to access representations of high frequency words at the
word level representations, it is expected that little activation at the component character level is
required to aid the access of high frequency words.
While studies on Chinese reading processing in adults have been extensively done (Zhang
& Peng, 1992; Taft & Zhu, 1995; Taft, Liu & Zhu, 1999; Peng, Liu & Wang, 1999), literature
review indicates that the developmental trend of the reading processing of disyllabic words in
Chinese has rarely been investigated. The aims of the present experiment are two folded: on top of
testifying the effects of whole word frequency and component character frequency, children’s
reading development and the change of the interaction between whole word frequency and
component character frequency across grades will also be investigated. A reading aloud task was
used and the percentage correct score of their reading performance was employed at the dependent
variable.
METHOD
Subjects
A pool of 120 children was obtained by picking 20 students (10%) above and 20 students
(10%) below the median calculated by their examination marks on Chinese Language from Grade
one, three and five. All subjects were born in Hong Kong and studied in local kindergartens. Each
of the 120 children were administered Raven’s Standard Progressive Matrices (Raven, 1986), a
standardised non-verbal intelligence test, and the Chinese word reading sub-test of the Hong Kong
Test of Specific Learning Difficulties in Reading and Writing (Ho, Chan & E.D., HKSAR
Government, 2000). Twenty-four children in each grade, a total of seventy-two children, who
obtained scores between the 25th and the 75th percentile in the two tests mentioned above were
selected to be the subjects of the present study.
Materials
In order to study the levels of reading processing of the school-aged children, three sets of
80 two-character words were selected as stimuli for subjects from grade one, three and five
respectively, from the Hong Kong Corpus of Primary School Chinese (Leung & Lee, 2002).
Words in the corpus were from the primary-school textbooks of the Chinese subject. The criteria
for word selection were based on (i) the whole word frequency and (ii) the component character
frequency in the corpus. The frequencies were calculated cumulatively in each grade. For example,
the cumulative frequency of a word in grade three was the sum of the frequency of that word in
grade one, two and three. Similarly, the component character frequency of a character in grade five
was the total frequency of that character, appeared individually or incorporated in compound
words in the corpus, in grade one to five. The word frequency and the component character
frequency calculated in this way represented children’s accumulative exposure to Chinese words
and characters since the start of primary school education.
The two-character words in each grade in the corpus were categorised into high frequency
and low frequency words. Twenty percent of the total words in each grade with the highest
frequencies were defined as high frequency words. All those words with a frequency of one, which
occupied 41% to 46% of the total words in each grade, were categorised as low frequency words.
The pool of characters in the corpus was also divided into high and low frequencies in each
grade. The criterion for high frequency characters in the corpus was the 20% of total characters
with the highest frequencies in each grade. The definition of low component character frequency
was obtained by adjusting the percentage of characters with the lowest frequencies in the corpus to
a level that there was just enough number of stimuli in each grade. There were two reasons for
defining the low component character frequency using this method. First, there were limited words
with low frequency component characters, especially in grade one. Another reason was to ensure
that the component character frequencies in the stimuli were sufficiently low. Consequently, 75%
of characters with the lowest frequencies in the character pool of grade one were defined as low
frequency, which was 1 to 21. Fifty-eight percent of the lowest frequency characters in grade three
and five in the corpus were categorised as low frequency, which was 1 to 16 in grade 3 and 1 to 22
in grade five.
There were four types of component character frequency constructions in the two-character
words: high-frequency first and second characters (HH), high-frequency first character and lowfrequency second character (HL), low-frequency first character and high-frequency second
character (LH), and low-frequency first and second characters (LL). Combining whole word
frequency with component character frequency, there were eight conditions of stimuli for each
grade. In each condition, ten words were selected randomly from the Corpus according to the word
frequency and component character frequency. Three lists of 80 two-character words were
generated for each of the three grades involved in the study.
Each stimulus was printed in black using the ‘biau kai’ (標楷體) font, sized 40 at the centre
of a plain white paper in one-fourth of an A4 size. The default character spacing of the font was
used.
Design and Procedures
The data collection was carried out individually in a classroom at the primary school that
participated in this study. Every subject was asked to read aloud the 80 stimuli designed for each
grade mentioned above. One word was presented at a time and there was no time limit for the child
to read the stimuli. The children were instructed to read the word stimuli as many as they can.
They were encouraged to guess when they were not sure or had no idea about the words. The
children’s responses were audio-taped and recorded online with a scoring form. The read aloud
task was repeated for 72 subjects and a total of 5760 word productions were collected.
Measurement
One mark was given for each correctly read character and zero mark was given for either
no response or incorrect production of the character. The reading productions of the subjects were
audio taped for double-checking and for transcription of the error productions in romanization.
RESULTS AND DATA ANALYSIS
The mean percentage correct in each frequency conditions of each subject was calculated.
The mean scores were then entered into a 3 x 2 x 4 three-way ANOVA with repeated measures.
The subjects’ grade was the between group factor, which had three levels: grade one, three
and five. The whole word frequencies and the component character frequencies of the stimuli were
the within group factors. The word frequency factor consisted of two levels, high and low word
frequency, while the component character frequency factor had four levels: HH, HL, LH and LL.
The mean scores of the subjects’ reading performance were the dependent factor.
In order to understand the developmental trend of the primary school students’ reading of
Chinese words, the means and standard deviations of the subjects’ overall percentage correct were
calculated. The reading performance in each grade in different word frequency and component
character frequency conditions was shown in Table 1.
Insert Table 1 about here
Result of the 3-way ANOVA with repeated measure revealed a significant main effect of
grade (F(2,69) = 10.31, p < .001) with the reading performance increased with grade (Table 3).
The differences between grades that contributed to the main effect were investigated with Tukey
HSD test. Grade three and five students were found to perform significantly better than grade one
students(p < .005). Since the frequency was controlled across grade, the results could not be
accounted for by the fact that higher graders have more exposure to characters. An alternative
explanation was that higher grades have acquired better reading skills when compared with the
lower grade.
Significant word frequency main effect was found (F(1, 69) = 300.01, p < .001), with the
reading accuracy of high frequency words better than low frequency words. Significant
component character frequency main effect was also found (F(3, 207) = 129.45, p < .001). Tukey
HDS test showed that all pair-wise comparisons between levels of component character frequency
were significant (p < .005) with HH > HL > LH > LL (Table 1). These findings agreed with Taft et
al. (1994)’s result as reported by Peng et al. (1999).
No significant interaction effect between grade and word frequency was found. However,
there was a significant interaction effect between grade and component character frequency (F(6,
207) = 14.66, p < .001) (see Figure 2). Tukey HSD test was employed to study the contribution of
each of the levels of the two factors in the interaction. Results showed that HH words were
significantly different from LL in all three grades (p < .001) indicating that all three grades read
words with HH words significantly better than words with LL component character frequencies.
No significant difference was found when reading accuracy of HH words across grades were
compared. However, in LL frequency condition, grade one students performed poorer than grade
three and five students (Table 1). The pair-wise comparisons showed that the significant difference
(p < .001) between grade one and grade three, and between grade one and grade five.
Insert Figure 2 about here
Significant interaction effect between word frequency and component character frequency
was also found (F(3, 207) = 70.06, p < .001). The interaction was further investigated by using the
Tukey HSD test. Results indicated that there was no significant difference among the levels of
component character frequency in high word frequency condition, whereas, significant difference
was found among the component character frequency constructions in low word frequency
condition (p< .001). Figure 3 illustrated that the subject’s performance decreased in the order of
HH > HL> LH > LL when the word frequency was low.
Insert Figure 3 about here
Three-way interaction among all factors was statistically significant (F(6, 207) = 3.99, p
< .001). The Tukey HSD test was carried out to investigate what contributed to the significant
interaction. No significant difference across grades could be found for HH words. Under HH
condition, the performances of students across grades, word frequency were not significantly
different. In contrast, under HL, LH and LL conditions, significant differences were found across
word frequencies and grades (p < .001).
In grade one, under high word frequency condition, performance on LL words were
significantly poorer than HH, HL and LH words (p < .001). The higher graders read words with
high word frequency similarly well, despite the variation in component character frequencies. In
the low word frequency condition, the difference between HH and LL component character
frequency was statistically significant in all three grades (p < .001). As shown in Figure 4, the
difference was largest in grade one and the difference decreased across grades.
Insert Figure 4 about here
Difference between HL and LH component character frequencies was not significant in all
word frequency conditions in grade one and three. However, grade 5 students performed
significantly poorer for LH words than HL words in the low word frequency condition (p < .001),
and LH words were not significantly different from LL words.
In summary, significant main effects of grade, word frequency and component character
frequency were found. Significant two-way interaction effects between word frequency and
component character frequency, grade and component character frequency, as well as 3-way
interaction of grade, word frequency and component character frequency were found. Two
interesting differences were found in the patterns of grade one and five. Children in grade one
performed worst on LL words when compared with HH, HL and LH words in the high word
frequency condition than other grades, while poorest performance for LH than HL words was
noted in the low word frequency condition in grade five.
DISCUSSION
Similar to the findings of previous studies (Zhang et al., 1992 & Peng et al., 1999) on word
level reading processing, the results of the present study showed that there was word frequency
effect in children’s reading. As mentioned in the introduction section (p.4), word frequency effect
on reading performance was expected if there was processing at word level as high frequency
words were more readily activated than low frequency words based on Taft et al.’s (1995)
interactive-activation model. The significant word frequency effect in the current study supported
the claim that primary school-aged children used word level processing in reading multi-character
words.
Main effect of component character frequency found in this study was also similar to
results obtained by Zhang et al. (1992) and Peng et al. (1999). If children used whole word level
reading processing only, the component character frequency would not affect their reading
performance, and high frequency component characters would not assist their compound word
reading. The presence of component character frequency effect implied that children used the
character level processing in reading multi-character words. It implies that school age children
analysed words into component characters and activated the character level representations in
reading.
The results obtained from the reading aloud task in this study indicates that School-aged
children are able to make use of representations at word level and component character level in
their reading. The question followed was in what way the word level and component character
levels of processing related to each other. Significant interaction effect between word frequency
and component character frequency was found in the present study. Significant component
character frequency effect with reading performance decreased in an order of HH>HL>LH >LL
was found in low word frequency condition. In contrast, Peng et al. (1999) found similar
component character frequency effect in high word frequency condition only.
The results of the present study were consistent with Taft et al’s (1999) multilevel
interactive-activation framework (Figure 1). When children read words with high word frequencies,
they could easily access the words at the word level and did not need to process the words into
their component characters. They read words with high word frequency equally well across
different component character frequency constructions. When they read low frequency words,
representations of the words were difficult to activate at word level. As a result, they processed the
words down a level – the component character level in order to read the words. In other words,
both the word level and component character level representations were activated in reading multicharacter words in children. The word level reading strategy was used for high frequency words
while the component character level reading strategy was applied for low frequency words.
It appeared that both the word level and the component character level awareness were
developed in children since grade one and they were also present in grade three and five. The
change of the reading strategy across grades was a matter of choice of the word and component
character levels. The reading strategies across grades were investigated by looking at the results of
the significant three-way interaction between word frequency, component character frequency and
grades. Students from all grades read high frequency words more accurately than low frequency
words for HL, LH and LL component character frequency conditions. The word frequency effect
was present in grade one, three and five, which meant that children in all grades used the word
level reading strategy in reading Chinese multi-character words.
Significant component character frequency effect was found in all grades in the low word
frequency condition, whereas only grade one student showed the significant effect of the
component character frequency under high word frequency condition. Students from grade three
and five read all high frequency words equally well. Results also indicated that children from
grade one, three and five made used the component character level reading strategy. Reading
performance of the grade one students was affected by the component character frequency in a
greater extent than grade three and five. It implied that grade one students had a higher tendency to
read Chinese multi-character words using the component character level reading strategy. The
higher graders tended to use the word level reading strategy more than the component character
level strategy, as their reading performance were less affected by the component character
frequency.
It was predicted that children’s reading performance for low frequency words with HH
component character frequency would be as low as other low frequency words. However, the
results showed that they read the words with HH component character frequency equally well in
both high and low word frequency conditions. Their success in reading low frequency words with
HH component character frequency could be explained by the processing of words at the
component character level when the whole word representations could not be found for the low
frequency words. It appeared that the word level and the component character level reading
processing compensated each other when the representation in either of the levels was not found
reflecting the flexible use of the processing levels of Chinese multi-character words by children.
The conclusions on the reading strategy used by different grades were further supported by
the error patterns produced by the subjects. Based on the level of processing mentioned above,
errors produced by the subjects were categorised into five groups: whole word level access errors,
character level access errors, errors involved both levels of access, no response and errors that
could not be categorised.
The whole word level access errors were assumed to be a result of errors occurred at word
level processing. It was hypothesised that error occurred when a semantically related word, which
was also activated by the target two-character word, was wrongly selected as the output. An
example of the whole word level access error was the activation of the word 規則 /khwi55 tsAk5/
meaning ‘rule’ in place of the target 秩序 /tit2 tsJy22/ meaning ‘order’, which were similar in
word meaning.
The character level access errors were errors resulted from the competition of the target
words and related entries at the component character level. For example, the target 精神 /tsiN55
sAn21/ which means ‘spirit’ was replaced by 清神 /tshiN55 sAn21/ which is a non-word, where the
first character was orthographically similar to the first character of the target word.
The third type of errors was generated in the processing involving both word level and
component character level. One example of this type of errors was the replacement of the target 友
愛 /jAu23 Oi33/ which means ‘friendship’ by 朋愛 /phAN21 Oi33/ which has no meaning, where 朋
友 /phAN21 jAu23/ is a word meaning ‘friend’. It was suggested that the child access the
representation of 友 and 愛 at the component character level. However, no representation of the
character 友 was found, so the subject activated words that contained the character 友 at the word
level representation and he found the high frequency word 朋友. The activation at the word level
was then sent down back to the component character level where mistake occurred. The character
representation of 朋 was activated instead of that of 友. As a result, the child produced 朋愛 in his
response.
Insert Table 2 about here
Table 2 showed that the numbers of whole word level access error in all three grades were
similar for low frequency words. Children in different grades made similar number of errors that
were generated during word level processing. It supported the word frequency main effect and the
non-significant interaction effect of grade and word frequency, which suggested that word level
reading strategy was developed in primary school-aged children and the word frequency effect was
constant across grades.
The number of character level access errors was the highest in grade one and decreased
across grades. This finding provided further evidence on the main effect of component character
frequency and its interaction with grade. Reading of grade one students were affected the most by
component character frequency and thus made more errors during character level processing.
The number of errors made in the process involving both levels of representation was also
the highest in grade one and decreased across grades. This trend was assumed to be due to the
component character frequency effect that occurred when the representation at the character level
was accessed. The result also supported the conclusion that grade one students used the component
character level access more extensively than higher graders, while higher graders tended to access
multi-character words at the whole word level and was less affected by component character
frequency.
If Peng et al.’s (1999) findings that component character frequency effect was found only
in high word frequency condition was correct, the component character reading strategy would be
activated at high frequency words level. Accordingly, More character level access errors would be
resulted for reading high frequency words than low frequency words. However, the fact that the
number of character level access errors was larger for low frequency words than for high
frequency words in the present study seemed to suggested the opposite. Evidence from the error
pattern in the current study further objected Peng et al.’s (1999) conclusion. It is likely that Peng et
al.’s finding is due to the use of lexical decision paradigm where subjects were “forced” to make a
binary on word.
A special finding was found in grade 5. The performance of LH component character
frequency in low frequency words was significantly lower than HL in low frequency words in
grade 5, while there were no differences between HL and LH component character frequencies in
other word frequency conditions and in other grades. The grade 5 students read words with LH
component character frequency as poor as LL in low word frequency. This finding disagreed with
the findings in the studies done by Zhang et al. (1992) and Taft et al. (1994), as reported in Peng et
al. (1999). To understand the phenomenon, the component character frequencies and word type
(compound word vs. binding word) of the stimuli used in the HL and LH conditions with low word
frequency in grade 5 were examined.
The reading processing of compound words and binding words were different. Taft & Zhu
(1997) suggested that characters from compound words, no matter whether they were free or
bound morphemes, could be accessed at both character and whole-word levels. In contrast,
according to Taft & Zhu (1995), characters from binding words were accessed at whole-word level
because they are cannot be independently used. Therefore, the binding words included in the
stimuli might contaminate the results. It was proposed that children would use word level
processing only when reading binding words, where the character level processing would be
inactivated.
The H and L component character frequencies of the stimuli in LH conditions were
comparable to that in other conditions, HH, HL and LL. The poorer performance in LH component
character condition was not affected by the deviation in the frequencies of the characters in the
stimuli. When looking into the types of words selected, there was no binding word in the stimuli of
LH. There was always a high frequency component character, which could combine with a number
of different characters to form other words, in either the first or the second position in the LH word.
Therefore, the poorer performance in LH condition was not due to the involvement of any binding
word. The finding could only be interpreted as the characteristic in the reading ability of the higher
graders. It was suggested that a position effect of the component character frequency was shown in
grade 5 when they read low word frequency words. It was hypothesized that high frequency
component characters in the first character position in words facilitated the activation of the whole
words at word level in low word frequency condition in higher graders. The LH construction had
no facilitative effect, or even inhabitation, on the activation of the whole word in the view that the
performance in LH condition was as low as in LL condition for low frequency words in grade 5.
The reading performance of children increased with grades, which reflected that the
children’s reading ability increased as they received longer education. The frequencies of the twocharacter words and their component characters were controlled for each individual grade in this
study to make the reading performances of the three grades comparable. This eliminated the
possibility that the increase in performance across grades was due to subjects’ increased exposure
to Chinese words and characters as they received years of education.
In conclusion, children as young as grade 1 read multi-character words using both whole
word level and component character level reading strategies. Lower graders’ reading performance
was affected by the component character frequency to a greater extent than higher graders’. The
lower graders are more dependent on the component character level reading strategy than higher
graders, while the higher graders tended to read Chinese multi-character words as meaningful
wholes.
Limitations
There were two limitations in this study. Firstly, due to the limitation of words available for
the selection of stimuli, binding words were also included in the stimuli with LL component
character frequency. As mentioned above, binding words were accessed at word level, as their
component characters were not single morphemes. It was considered that the present of binding
words in the stimuli decreased the component character frequency effect to a certain extent.
Secondly, the source of the stimuli was developed from the primary textbooks of the
Chinese Language subject only. The variable that a child learned words and characters in other
subjects’ textbooks, leisure books and newspapers was not controlled. The frequencies and number
of different words and characters in the corpus could not entirely reflect children’s real exposure to
Chinese characters and words. Some of the low frequency words and component characters used
in this study might not be really unfamiliar to children who did lots of readings. Nevertheless,
words and characters obtained from the Chinese Language textbooks could be an index of the
primary school-aged children’s repertoires of orthography, as children learned most words in
Chinese Language lessons.
Further studies
In order to obtain more representative results in future developmental studies on reading
Chinese, it was recommended to include words and characters in textbooks of other school
subjects in the selection of stimuli. This ensured the stimuli could reflect the children’s real
exposure to Chinese characters and words in different areas and contexts.
Further studies on the characteristics of multi-character words in children’s textbooks were
suggested to investigate the types and quantity of words learned by children across grades. The
results could be used to study the effect of children’s exposure to different types of words on the
changes in reading strategies across grades.
Studies on the development of reading strategies in pre-schoolers were recommended.
Since the results of the present study revealed that children as early as grade 1 had already
developed word level and character level reading strategies. The development of these readingprocessing levels was suggested to happen before children entered primary schools.
Finally, as positional effect of component characters on the reading accuracy of low
frequency words was found in grade 5, investigation on the effect of HL and LH component
character frequency constructions on reading in higher graders was suggested. This helped us
understand whether the positions of high and low frequency component characters could facilitate
the reading of two-character words.
Implications
The findings of the present study showed that higher graders tended to read two-character
words as whole units while lower graders had less advanced word level awareness in reading.
Considered that the content of the primary textbooks were passages, children was assumed to read
the passages character by character before they realised or learned the words in the passages. This
could explain why lower graders’ word awareness was lower. It is advisable for teachers foster the
development of both word level and character level awareness and the flexibility in using different
reading strategies in primary school-aged children.
REFERENCES
Caramazza, A., Laudanna, A., & Romani, C. (1988). Lexical access and inflectional morphology.
Cognition, 28, 297-332.
Ho, C. S-H., Chan D. W-O. & Education Department, HKSAR Government. (2000). The Hong
Kong Test of Specific Learning Difficulties in Reading and Writing. HK: E.D., HKSAR
Govt.
Leung, M-T. & Lee, A. (2002). The Hong Kong Corpus of primary School Chinese. Paper
presented at the 9th Meeting of the International Clinical Phonetics and Linguistics
Association. Hong Kong.
Liu, Y. & Peng, D-L. (1997). Meaning Access of Chinese compounds and its time course. In H-C.
Chen (Ed), Cognitive processing of Chinese and related Asian Languages. Hong Kong:
The Chinese University Press
Peng, D., Liu, Y. & Wang, C. (1999). How Is Access Representation Organized? The Relation of
Polymorphemic Words and Their Morphemes in Chinese. In J. Wang, A.W. Inhoff & H-C.
Chen (Ed), Reading Chinese Script: A Cognitive Analysis. New Jersey: Lawrence Erlbaum
Associates.
Raven, J.C. (1986). Hong Kong Supplement to Guide to the Standard Progressive Matrices. H.K.:
Education Department, H.K.
Taft, M. (1994). Interactive-activation as a framework for understanding morphological processing.
In D. Sandra & M. Taft (Eds.), Morphological structure, lexical representation and lexical
access. East Sussex, U.K.: Lawrence Erlbaum Associates.
Taft, M., Liu, Y. & Zhu, X. (1999). Morphemic processing in reading Chinese. In J. Wang, A.W.
Inhoff & H-C. Chen (Ed.), Reading Chinese script: A cognitive analysis. New Jersey:
Lawrence Erlbaum Associates.
Taft, M. & Zhu, X. (1995). The representation of bound morphemes in the lexicon: a Chinese
study. In I.B. Feldman (Ed.), Morphological aspects of language processing. Hillsdale,
N.J.: Lawrence Erlbaum Associates.
Taft, M. & Zhu, X. (1997). Using masked priming to examine lexical storage of Chinese
compound words. In H-C. Chen (Ed.), Cognitive processing of Chinese and related Asian
languages. HK: The Chinese University Press.
Zhang, B. & Peng, D. (1992). Decomposed storage in the Chinese lexicon. In H.C. Chen & O.J.L.
Tzeng (Eds.), Language processing in Chinese. Amsterdam: Elsevier Science Publishers.
Zhou, X., Shu, H., Bi, Y. & Shi, D. (1999). Is There Phonologically Mediated Access to Lexical
Semantics in Reading Chinese? In J. Wang, A.W. Inhoff & H-C. Chen. (Ed.), Reading
Chinese Script: A Cognitive Analysis. New Jersey: Lawrence Erlbaum Associates.
Figure 1. The multilevel interactive-activation framework illustrating the reading process using the
example of the word 現代 (adopted from Taft, Liu & Zhu, 1999)
Table 1
Mean percentage correct of the subjects’ overall performance and that in different word and
component character frequency conditions
Overall percentage
score (%)
Mean percentage score (%)
Word
frequency
High
Low
Component character frequency
HH
HL
LH
LL
99.80
94.40
93.15
80.20
95.20
78.55
77.10
54.60
90.68 (5.48)
Word
frequency
High
Low
99.40
97.10
97.30
85.20
95.00
83.15
99.40
68.95
92.21 (3.90)
Word
frequency
High
99.60
98.95
98.15
98.35
Low
99.60
89.60
77.10
76.45
Grade
grade 1
Mean (S.D.)
84.11 (9.17)
grade 3
grade 5
100
Percentage correct
90
Grade 1
80
Grade 3
Grade 5
70
60
HH
HL
LH
LL
component character frequency
Figure 2. Interaction between Grade and Component character frequency
Percentage correct
100
90
High word frequency
80
Low word frequency
70
60
HH
HL
LH
LL
Constituent character frequency
Figure 3. Interaction between Word frequency and Component character frequency
Table 2
Number of different errors in high and low word frequency conditions in grade 1, 3 and 5
Word frequency
Whole word level access errors
Character level access errors
Errors involved both levels of access
Other errors
No response
Total errors
Grade 1
High
Low
25
42
51
172
4
23
6
25
30
76
454
Grade 3
High
Low
16
55
22
100
0
20
2
36
0
40
271
Grade 5
High
Low
6
64
7
116
0
9
7
40
1
15
255